Vertex AI | Gemini | Generative AI - GCP Machine learning

Create a Google Cloud free trial account to access $300 credit for 91 days, sign in with your Google account, and provide your country and billing details to start.

Explore how Google Cloud regions and zones enable low latency and high availability by deploying across multiple data centers and regions, supporting disaster recovery.

Create a google cloud service account to enable programmatic access to google cloud APIs via python. Generate a json key for authentication and use it in code.

Set up Vertex AI code in a Jupyter Lab notebook by loading the service account key, configuring project ID and region, and initializing Vertex AI to access generative AI APIs.

Explore text generation with Gemini models in Vertex AI by creating a generative model, triggering generate_content on Gemini 2.0 Flash or 1.5 Flash, and experimenting with streaming mode.

Learn to generate text with Vertex AI API using Gemini 1.5 Flash, configure generate_config parameters like temperature, top_p, top_k, max tokens, and explore memory-enabled chat with start_chat and history.

Create a reusable module for boilerplate code in Vertex AI projects by importing myvertexai.py into new notebooks and running it from a conda environment in Jupyter Lab.

Explore prompt engineering principles to improve results, focusing on concise, specific prompts, well defined prompts, and one question at a time to guide LLMs.

Learn to control model hallucination with temperature and knowledge cutoff awareness, test prompts like P1 and P2, and apply one-shot and few-shot prompting for classification in Vertex AI.

Learn to convert text into semantic vector representations with Vertex AI text embedding API, select a pretrained 005 model, and compute cosine similarity using numpy dot products.

Generate combined text embeddings for multiple items and sentences, revealing three 768-d vectors for banana, apple, and cat. Compare English and multilingual models using cosine similarity to understand model choice.

Explore how Vertex AI enables task-specific embeddings by specifying the task type (semantic similarity, classification, question answering, code retrieval, clustering) using TextEmbeddingInput and TextEmbeddingModel to improve vector search.

Learn batch data embeddings at scale by placing input jsonl in Google Cloud Storage, outputting embeddings to BigQuery via textembedding_model.batch_predict and defining input_uri and output_uri.

Set up a BigQuery dataset and input table, run batch data embedding with Vertex AI, and store embedding outputs back to BigQuery for retrieval documents.

Execute batch data text generation with Gemini on Vertex AI, using a JSONL input from a Google Cloud Storage bucket and saving results to a defined output URI.

Learn to perform batch data text generation with Gemini on BigQuery, using a Gemini table and input_uri and output_uri for end-to-end batch prediction.

Build multi-model pipelines that combine image and text prompts in Vertex AI Gemini models, supplying images via uri and mime_type to generate image-based insights like fruit recognition and nutrition.

Master multimodal processing in Vertex AI by using multiple images, describing both with image_file1 and image_file2, and noting differences like a free-form stack versus a structured pyramid.

Learn how to perform multimodal video processing inside Gemini and Vertex AI, sending video and audio prompts, describing video content, and summarizing audio in a single request.

Master document processing with Vertex AI and Gemini LLM, loading Part and GenerativeModel, and summarizing PDFs or text via MIME types.

Navigate the Google Cloud console to explore Vertex AI Studio, a prompt-driven testing and deployment suite for generative AI, featuring freeform and chat modes with prompt galleries and management.

Explore Vertex AI Studio as a testing bed for prompts, generating code and curl outputs. Adjust prompts and token limits, attach media, and test image and video inputs for summarization.

Explore Vertex AI Studio through a console walkthrough, learning to use Gemini flash and Imagen 3 models, manage prompts, adjust temperature and tokens, and perform translation, text-to-speech, and speech-to-text tasks.

Explore how to generate high-quality images with a Gemini class model in Vertex AI, using Imagen 3.0 and fast variants, via zero-shot prompts and saving outputs.

Generate code with the Google code gecko model in vertexai, using a low temperature for deterministic output. See loading the model, predicting with a prefix, and python examples like A+B.

Explore function calling in Gemini to extract currency conversion details from user prompts, call a rate API with from, to, and date parameters, and generate precise, current results.

Explore model garden in Vertex AI Studio to access a diverse collection of models, filter by modality and task, and deploy via Hugging Face or API services within Google Cloud.